This thesis concerns user modelling and artificial intelligence in the context of supporting users in their day-to-day kitchen activities (basically cooking). Due to cooking being one of the most common tasks that many persons have to face every day, it is important to develop some kind of computer system that helps the users. This system should help the users to decide what to cook, how to cook and what to use for cooking, suggesting food items or recipes according to their personal features (such as skill, taste or time available, for instance).
In this report we describe and summarize the process of design and creation of an experimental decision engine for a system that works as a kitchen assistant. This engine can learn about the users’ personal features and recommend recipes or food items that match with these users. The purpose is to make it easier for them to decide what to cook every day. This software can be especially useful for people with dietal restrictions, young students or elderly people, but everybody can take a clear benefit from its use.
The software developed in this project was tested with real users in order to determine how good it was. The results were at least acceptable, when not very suitable, even with some intrinsic limitations in certain aspects (like the limited content of the prototype databases). These results showed that the recommendations of the program were more appropriate for the users when more (and more complete) information about them was used. Some extra features of the program oriented to help the users with the steps of the recipes and with the navigation in the program also gave positive results.